Facebook published its first community standards report detailing the actions it has taken on content that contains violence, sexual material, terrorism-related content, hate speech, and spam messages in an ongoing attempt to restore user confidence in the social network.
The report comes a few weeks after Facebook first published detailed internal guidelines on how to implement content removals.
The report was written by Guy Rosen, vice president of product management, with Alex Schultz, vice president of data analysis at Facebook. “Facebook deleted 583 million fake accounts in the first three months of 2018. This number of counterfeit accounts is Less than the number of false accounts that were deleted in the fourth quarter of 2017, which amounted to 694 million accounts, and this does not include millions of counterfeit accounts discovered by Facebook before the completion of the registration process and delete them.
These figures give users a better idea of the sheer volume of counterfeit accounts that Facebook handles. In recent months, the company has pledged to use facial recognition technology to detect fake accounts that may use someone else’s image as a profile image.
But a recent Washington Post report found that face recognition technology from Facebook may be limited when it comes to detecting fake accounts because the tool does not yet scan the image against all images published by all 2.2 billion users when searching for Counterfeit Accounts.
Facebook also provided an analysis of spam that was removed during the first quarter of 2018, as well as the amount of data reported by its systems or reported by users:
- 837 million spam was deleted in the first quarter of 2018 – 100% of which was found before users had a chance to report it.
- 21 million publications of the content were stripped of inappropriate sexual activity, 96% of which was discovered first by Facebook tools.
- 3.5 million publications of violent content were deleted , of which 86 % were identified by Facebook tools.
- 2.5 million hate-hate publications were deleted , 38% of which were found using Facebook tools based on artificial intelligence techniques.
These figures show that FB is still mostly dependent on other people to disclose the hate speech – which CEO Mark Zuckerberg has spoken about before, saying it is much more difficult to create an artificial intelligence system that can identify hate speech. Where Facebook defines hate speech as “a direct attack on people based on race, nationality, gender, religious affiliation, sexual orientation, disability or illness.”
As Mark Zuckerberg said at the F8 Developers Conference, we have a lot of work to do to prevent abuse. On the other hand, artificial intelligence techniques are promising, but they are still far from being effective for most of the bad content because understanding the context is very important.
Facebook confirms that it will continue to provide updated figures every six months. The report, which was published today from October 2017 to March 2018, has a rating comparing the volume of content that the company has taken action on in different categories in the fourth quarter of 2017 and the first quarter of 2018.